Artificial intelligence approach to predict the penetration and softening point of graphene oxide modified asphalt

Abstract

Penetration and softening point are the two most important criteria for classifying asphalt grades according to penetration. The determination of these two parameters of modified asphalt graphene oxide (GO) by
experimental method encountered certain difficulties due to the high cost of GO and long experimental time. The purpose of this study is to use the adaptive neuro-fuzzy inference system (ANFIS) combined with the genetic algorithm (GA) to predict the penetration and softening point of GO modified asphalt. Two datasets including the penetration dataset (122 samples), softening point dataset (130 samples) collected from 12 different studies with 9 input parameters, are used to construct and test the data digital simulation tool. In addition, the study uses a 10-fold cross-validation technique along with statistical criteria such as correlation coefficient (R) and root of mean square error (RMSE) to evaluate the performance of the models. The research results show that, for the penetration dataset, the ANFIS-GA model has RMSE = 6.045 (0.1 mm), R = 0.949, the ANFIS model has RMSE = 8.492 (0.1 mm), R = 0.893. For the softening point dataset, the ANFIS-GA model has RMSE = 1.848 (oC), R = 0.991, the ANFIS model has RMSE = 13.863 (oC), R = 0.818. This shows that both ANFIS-GA and ANFIS models have good predictive performance and high accuracy. With smaller RMSE and higher R in both datasets, the ANFIS-GA model is evaluated to be better than ANFIS. This model can completely be applied to help materials engineers save time and experimental costs.